The Unfinished Business of GSE Systemic Risk: Mortgage Credit Risk Concentration Is Getting Worse, Not Better (Part 2 of 2)

Introduction  

The 2008 Great Financial Crisis (GFC) was the most severe episode of systemic financial risk since the Great Depression. In response, the government enacted legislation, amended existing regulations, and introduced new ones, all in an effort to prevent a similar collapse in the future. However, the two government-sponsored enterprises (GSEs), Fannie Mae and Freddie Mac (F&F), were largely excluded from those actions despite their near-failure being a major contributor to the GFC. [1]  This exclusion, however, made sense because the Obama administration’s policy at that time was for Congress to develop a replacement for F&F, after which the two companies would be wound down and go out of business.[2]

However, Congress never delivered such a replacement. Even now, 17 years later, none is in sight. As a result, F&F remain the largest source of American residential mortgage credit for the foreseeable future. Unfortunately, the government never went back to address how to remediate their ability to contribute towards financial instability and systemic risk, leaving a significant gap that still persists today. 

This two-part series examines those unresolved GSE systemic risk issues. In Part 1, the three sources of major potential instability at F&F  — large enough to possibly trigger a systemic risk event —  were identified.  The discussion then described how two of those sources — the excessive concentration of mortgage interest rate and liquidity risk, and F&F’s severe undercapitalization — have already been successfully contained.[3]

Part 2 discusses the third source of such instability: the excessive concentration of mortgage credit risk. Unlike the first two sources, this risk has absolutely not been contained. Even worse, it is actually going in the wrong direction, with the risk currently reconcentrating back to very high levels.  

As background, in response to the very large losses by F&F from the credit risk on their mortgages, the FHFA in 2012 challenged F&F to develop a method to transfer credit risk to others. The result, known as the GSE credit risk transfer (CRT) program, is the primary focus of this article.[4] 

Commencing in 2013 with the completion of the first-ever GSE CRT transaction, the program steadily reduced F&F’s excessive concentration of mortgage credit risk by transferring large amounts of it to a diversified global pool of institutional investors and insurers. Moreover, this was done on an economically efficient basis, i.e., at no overt economic cost to the two companies or the government.[5] As a result, a massive reduction in concentrated GSE credit risk appeared to be inevitable[6] as old mortgages without associated CRT were paid off or matured and replaced by new ones that included it.

However, starting in mid-2019, the FHFA, under new leadership, unexpectedly began to signal that it no longer viewed the CRT program positively. By 2021, the CRT program was visibly becoming hollowed out. While the program still operated, a new capital rule adopted by the FHFA at the end of 2020 has directly led to less and less risk being transferred away from F&F. It is unclear how much of this hollowing out was intended by the FHFA versus being an unintended consequence that still continues.  Regardless, credit risk has been reconcentrating ever since.  

Part 2 of this series urges the FHFA to once again focus on significantly reducing the excessive concentration of mortgage credit risk at F&F by restoring the strong, comprehensive CRT program that existed from 2013 to 2019. If the FHFA is unwilling to do this on its own, the Financial Stability Oversight Council (FSOC), established after the GFC to address just such issues, should leverage its influence and authority to help the FHFA recognize the importance of doing so.

CRT from 2013 to 2019: A challenge quickly becomes a policy success

Since the 1980s, F&F’s business model has relied on funding its loan purchases through securitization. The two companies guarantee that investors who buy their pass-through mortgage-backed securities (MBS) will not suffer a credit loss. As a result, well over $6 trillion of single-family first mortgage credit risk – about half the total outstanding nationwide – is today concentrated in these two companies.  

This is a tremendous source of potential financial instability and systemic risk if ever there was one. Looking back, it was large losses on their mortgage assets and guarantees, and the likelihood of more to come, combined with their historic severe undercapitalization, that caused their near-failure in 2008 when the market lost confidence in them.[7]

In early 2012, the FHFA released its first annual “conservatorship scorecard,” which directed the GSEs to do a variety of things to make the companies and mortgage markets work better. Among these was a challenge for the GSEs to develop a credit risk transfer program for single-family mortgages.[8] There were several potential benefits if such a program could be successfully implemented on a large scale.[9] One was to directly reduce their massive systemic-level concentration of mortgage credit risk which had been so problematic during the GFC.[10]

The conservatorship scorecard’s CRT challenge actually encompassed multiple objectives and constraints, with three in particular standing out as most noteworthy. Specifically:  

• It absolutely had to leave undisturbed the operations and structure of the MBS markets into which F&F issued.[11] That means the credit risk guarantee given to MBS investors had to remain in place, leaving F&F fully responsible for any credit losses; risk transfer would then operate by having F&F develop a mechanism to transfer away part or all of those losses. 
• It would not cause mortgage rates to go higher. A CRT program had to be structured very carefully so that potential CRT investors — whose memories of the large mortgage credit losses during the GFC remained fresh in 2012[12] — would still feel comfortable investing in the asset class without demanding above-market returns. Otherwise, such above-market returns could require that the GSEs pay out so much to transfer the credit risk that mortgage rates would likely need to increase.  
• It had to transfer away specifically the risk of large losses that, during a period of significant stress, might cause the GSEs to become financially unstable. This meant transferring away the risk not of the routine losses that the GSEs typically absorb through their quarterly credit loss provisions, but the much larger ones that have the realistic potential to occur during a period of significant market distress.  

In response to the FHFA’s challenge in the scorecard, Freddie Mac — after more than a year of development — conducted the first-ever GSE CRT bond transaction in July 2013.  Fannie Mae followed later that year, using the same structure. This initial transaction, designed roughly like an insurance-linked note[13]— by which CRT investors agree to reimburse F&F for a specified set of credit losses — set a pattern for almost all CRT transactions since then.[14] Importantly, an insurance contract version of the structure was developed fairly quickly to complement the CRT bond version, enabling access to risk capital at large insurance companies.  

The result was an almost-immediate success, and the program quickly gained a reputation for operating well on a large scale through 2019. More specifically, the CRT track record during those years showed the following:

• CRT quickly became standard practice in most single-family MBS issuance. In 2013, the first year of the CRT program, the mortgage pools backing CRT transactions by F&F totaled $76 billion; this more than quadrupled to $345 billion in 2014 as the program was ramping up. By 2016, the FHFA’s annual conservatorship scorecard for the first time required that 90 percent of the principal of new “targeted”[15]mortgages have CRT behind them. As Table 1 below shows, during the years of the fully robust CRT program — which includes CRT transactions from 2015 to 2019, linked to MBS issuances from 2014 to 2018[16] — about 60 percent of mortgage principal (including both targeted and not) financed each year was associated with CRT.[17]
• The risk transferred during the typical CRT transaction was squarely in the area of the larger losses that had the most potential to cause a systemic risk event. An estimate of the routine losses (known technically as “expected losses,” and measured on a cumulative basis for the life of the underlying mortgages) was 0.15 to 0.30 percent.[18]  Thus, also as per Table 1 below, CRT transactions were definitely aimed at larger-than-expected losses by having deductibles][19] around or slightly above the expected loss range, specifically 0.10 to 0.40 percent. They also had a cap or maximum on total payouts for cumulative losses, set at around 3.50 to 4.50 percent; this resulted in CRT investors promising to reimburse F&F for losses up to about 3 to 4 percentage points above expected losses, which covered even the worst historical scenario.[20]
• The amount of capital needed to carry the credit risk on the pool of mortgages behind a particular CRT transaction was significantly reduced. The percentage reduction in capital required has not been specifically disclosed by the FHFA or F&F for individual CRT transactions completed over the years, but can be reasonably approximated from publicly available transaction details as being in the 50 to 60 percent range (also see Table 1). This is well short of 100 percent; the implication of this shortfall is discussed below.  
• CRT transactions were done on an economically efficient basis. This means that the amount paid to CRT investors, which comes out of the guarantee fee (G-fee) charged by F&F to primary market lenders, was low enough that the “implied cost of capital” on the CRT bonds and insurance contracts was lower than, or equal to, the cost of capital required for credit risk that F&F would retain.[21]This “economic efficiency” requirement was met for all of Freddie Mac’s CRT transactions[22] during the 2014 to 2019 period (as measured by Freddie Mac’s homegrown capital system prior to the FHFA adopting an official one in 2017, and by the official one after that).[23]

Viewed overall, then, the CRT program transferred a very substantial portion of the risk of larger credit losses, making a systemic risk event due to credit losses much less likely — though not entirely eliminated.[24] Thus, F&F’s business model  — which had historically been to buy and hold all of its credit risk despite the resulting massive concentration of that risk — became a hybrid: use CRT when it could be obtained on an economically efficient basis; otherwise retain the credit risk on a safe and sound basis by having a proper level of capital to carry it.  

It is important to note that the official government stress tests of F&F during those peak CRT years from 2015 to 2019 showed declining losses, reaching a relatively low level even under the “severe adverse” economic scenario specified by the Federal Reserve.[25] Therefore, considering (1) F&F’s more conservative credit risk appetite after the GFC,[26](2) the fortuitous upward trend in house prices that coincidentally began around 2012-2013, and (3) the actual CRT transactions booked during those years, it certainly appeared that F&F faced little risk of incurring the very large mortgage credit losses that could imperil financial system stability. In that sense, at least for that time period, the concentration of credit risk was fully contained, just not by CRT alone.

CRT from 2020 to today: The hollowing out

In the spring of 2019, Mark Calabria, a long-time critic[27] of the GSEs, became the new FHFA director. This triggered a series of events that resulted in the hollowing out of the CRT program, which continues to this day.  

This hollowing out derived from three key events. In chronological order, they were:

• First, the FHFA shifted its strategic direction. Director Calabria was also a CRT skeptic, stating his belief to me and others that while CRT might be effective during conservatorship, it would not be afterwards.[28]This viewpoint was officially reflected in the 2020 conservatorship scorecard, the first produced under his leadership, which downgraded the longstanding requirement to transfer credit risk on 90 percent of targeted mortgages by removing any mention of a numeric goal. While this step seemed minor or symbolic, it’s notable that Fannie Mae essentially stopped conducting any CRT transactions for part of Director Calabria’s more than two-year[29] tenure, which of course led to a reduction in the amount of risk being transferred, as explained below.

More importantly, during Calabria’s directorship, the FHFA issued at the very end of 2020 a new regulatory minimum capital requirement called the Enterprise Regulatory Capital Framework (ERCF), which naturally replaced the CCF for F&F to use in making their economic calculations.[30] It was controversial at that time for having inadequate risk sensitivity[31] and being overly complex. Both of these weaknesses proved to be central to the hollowing out of CRT.  

• Second, the ERCF systemically underestimated how much capital relief was delivered by CRT transactions. A key feature of a high-quality regulatory capital system is that its calculation of required capital, in response to individual transactions that take on or shed risk, moves up and down by an amount that credibly reflects the actual economic risks involved. Unfortunately, the inadequate risk sensitivity of the ERCF specifically resulted in it underestimating the amount of capital relief delivered by CRT transactions.  This, in turn, means that proposed CRT transactions looked uneconomic to pursue as the reduction in required regulatory capital was too small — due to the underestimate — to compensate for the portion of the G-fee that needed to be paid to CRT investors. The result, unsurprisingly, and in line with Director Calabria’s CRT-skeptic viewpoint, was a decline in the number of CRT deals completed.[32] As a result, as shown in Table 2, the percentage of securitized mortgages with CRT issued in 2020 dropped to 45 percent from its previous multi-year range of around 60 percent and then declined further to 36 percent the following year; such low levels had only last been seen when the program was brand new and ramping up back in 2014. This reduction reflected the first phase of the hollowing out of CRT.[33]   
• Third, the ERCF’s complexity unintentionally incentivized F&F to base their CRT transactions on extremely high deductibles, which deliver little if any actual risk transfer and thus also cost very little. Specifically, the ERCF included an unusual adjustment of the calculation of required capital to reflect house prices growing much more strongly than long-term historic levels.[34] This unintentionally allowed the GSEs to get capital relief on very high-deductible transactions (e.g., starting at 1.75 percent cumulative loss) at the same level as that of a much lower, more traditional deductible level (e.g., around 0.30 percent), which is clearly economically inaccurate. But such capital relief seemed very cheap, as CRT investors charged little for it since the actual economic risk being transferred was so small.  

Thus, as shown in Table 2 below, there was a second phase of the hollowing out beginning in 2022. F&F once again entered a higher volume of CRT transactions, rising from 36 percent to 53 percent of the associated MBS. But this appearance of greater risk transfer was deceiving, as the actual CRT transactions had increasingly high deductibles beginning in 2022 (rising steadily from the 0.30 percent average range in previous years to 1.78 percent just three years later), as shown in Table 2. That produced, unsurprisingly, only about one-third of the risk transfer impact found in pre-2020 transactions;[35] in fact, because the deductibles were so high, it is not clear whether such CRTs would actually generate any risk transfer dollar payments to F&F under any likely stress scenario. 

This means the CRT program today focuses more on what is sometimes called “capital arbitrage” than on real risk reduction. Money is spent mainly to create the appearance of major risk reduction, not the actual thing, because the ERCF is so misleading about the true underlying economics. The outcome: the impact of the CRT program has been significantly diminished.  

Conclusion and recommendation

From 2013 to 2018, the MBS issued by F&F experienced significant risk reduction via CRT transactions that occurred roughly a year later. These have now mostly matured and been replaced by MBS issued since 2019, which have little risk reduction from CRT. As a result, the concentration of credit risk at F&F is once again increasing. It remains unclear to this day how much this may be an intended outcome or is instead an unintended side effect that has not been adequately remediated through revisions to the ERCF. Regardless, this serves as a great example of a policy that seems to have backfired, turning a potential systemic risk reduction victory into a defeat. 

It is well past time for ERCF remediation to occur so the CRT program can once again be robust and based upon real-world economics. The FHFA needs to implement the necessary changes in ERCF — which will in some cases go beyond “tweaks”[36] — to restore the type of proper CRT program that existed before 2020. If it does not do so on its own, then the FSOC should step in, as it was designed to do, and use its influence and authority to ensure that the FHFA takes action.[37] Without such remediation occurring one way or another, the potential for the GSEs to create a systemic risk episode will remain very real.  

Footnotes

[1] To prevent their failure, the Federal Housing Finance Agency (FHFA), their regulator, placed them into conservatorship, naming itself as their conservator, and Treasury entered into a formal agreement to financially support their creditworthiness.

[2] This “wind-down” approach was the explicit policy of the Obama administration through its end in January 2017.

[3] While their containment has been successful, it is not necessarily long-lasting or permanent, as backsliding has happened in the past and could easily happen in the future.

[4] The CRT program has several significant benefits, of which systemic risk reduction is just one. For a discussion of this and the program more generally, see my three-part series “Demystifying GSE Credit Risk Transfer,” published in 2020. Part I – https://www.jchs.harvard.edu/sites/default/media/imp/harvard_jchs_gse_crt_part1_layton_2020.pdfhttps://www.jchs.harvard.edu/sites/default/media/imp/harvard_jchs_gse_crt_part2_layton_2020.pdfhttps://www.jchs.harvard.edu/sites/default/media/imp/harvard_jchs_gse_crt_part3_layton_2020.pdf.

[5] The definition of “economically efficient” is described later in this article.

[6] The time it would take was uncertain, as it depends upon future market conditions, especially those that would cause refinancing volumes to be high or low. From 2013, a rough estimate is that it would take at least 10 years, and more likely about 15, for CRT to reach its full effectiveness. That means it would have been fully effective roughly about now if it had continued as originally intended.

[7] That loss of confidence had to be very severe, as the companies benefited at that time from government support in the form of what was known as an “implied guarantee,” as explained in Part 1.

[8]See the 2012 Conservatorship Scorecard, page 2, “Risk sharing.” (https://www.fhfa.gov/document/ExecComp3912F.pdf).

[9] An example of such an additional benefit is to improve the market discipline on F&F, as the credit quality of the mortgages they securitized would, for the first time, become subject to a “second opinion” by CRT investors via what pricing they quoted for specific CRT transactions. For the full range of benefits, see the three-part series “Demystifying GSE Credit Risk Transfer” referenced above.

[10] Thus, it was thought that CRT would ideally do for mortgage credit risk just what the pass-through MBS had earlier done with the GSEs’ concentrated interest rate and liquidity risk, i.e., pass almost all of it through to a diversified global pool of large financial institutions, thus deconcentrating it.

[11] This included, in particular, a feature called the “to-be-announced” (TBA) market, which enhanced the liquidity of 30-year MBS markets enough that it traded at unusually low interest rates versus similar bonds. (A senior Treasury official once told me, upon my arrival at Freddie Mac in 2012, “Whatever you do, don’t do anything to mess up the TBA market.”)

[12] As a broad generalization, elevated mortgage credit losses from the GFC did not end until 2011 to 2012, as delinquencies only clearly declined from their high GFC levels starting roughly in 2012.

[13] Insurance-linked notes, commonly known as catastrophe (or “cat”) bonds, have operated without controversy since the 1990s.

[14] Many improvements were made in the initial structure during subsequent years to better align with tax, accounting, investor appetite, and other requirements, but the fundamentals remained unchanged.

[15] “Targeted” was defined by the FHFA as: (1) the loan-to-value ratio was above 60 percent, and (2) the interest rate was fixed for at least 20 years (which is dominated, of course, by the 30-year fixed rate mortgage). Under 60 percent loan-to-value ratio (LTV) loans were regarded as having such minimal risk that it was not worth F&F’s effort to transfer the risk. While few purchase loans fit this category, refinancing loans often did in the period of rising home prices that began about 2013.

[16] CRT transactions on the pool of mortgages behind MBS transactions occurred on average three-plus fiscal quarters after MBS issuance. In this paper, this is rounded to one year.

[17] The 60 percent calculation is based on the principal of the mortgages, not their riskiness. Because the largest group of excluded mortgages from the “targeted” pool had less than 60 percent LTV, the proportion of mortgages — adjusted for risk — that included CRT would be considerably higher than 60 percent.

[18] Source: Milliman, a well-known firm that provides actuarial and related risk management solutions. Milliman’s mortgage solutions group provides analytic software for CRT market participants.

[19] In the interest of making this article more accessible to a general audience, I avoided using the technical language employed by specialists in CRT. For example, deductibles, which almost everyone is familiar with from homeowner or auto insurance policies, are used to describe what are actually called “attachment points,” which are very similar; ditto for maximum payouts, which are actually called “detachment points.” The paper also does not delve into the even more technical field of tranches in CRT transactions, as it is beyond its scope.

[20] The losses incurred by F&F during the GFC are estimated to be from 3 to 4 percent (depending on whether one includes or excludes losses on certain riskier products no longer offered). That was the largest historic episode of mortgage losses since the Great Depression almost 100 years ago.

[21] The implied cost of capital mainly equaled the amount paid away to CRT investors (plus other related expenses) divided by the reduction in the capital required to be held by F&F because the CRT investors were absorbing larger-than-expected loss risk (and thus reducing the need for capital). Also, the FHFA, starting circa 2017, calculated an official “cost of capital” for F&F to use in their internal calculations each year; it ranged between 9 and 10 percent.

[22] Excluding some FHFA-mandated experimental ones.

[23] Freddie Mac’s homegrown capital system was very similar to what the FHFA eventually declared as the official one to use in conservatorship starting in 2017, called the “conservatorship capital framework” (CCF).

[24] One could argue that the FHFA or Treasury should have provided an economic incentive for F&F to transfer even more risk to achieve the policy goal of almost no potential for a systemic risk event, but they never did. Alternatively, the FHFA could have mandated that almost all credit risk be transferred away, which would have necessitated F&F “paying up” to get rid of more risk; this, in turn, would have driven up G-fees and thus mortgage rates, which was obviously undesirable to policymakers.

[25] The stress test results as of yearend 2015 were a loss of $49.2 billion; by 2019 yearend, they were down to $7.1 billion, which is quite a small amount given the size of F&F’s balance sheets. (Both numbers are the “without re-establishing a valuation reserve for deferred taxes” version, as that is the more relevant amount when losses are not large.)

[26] Much of this reduction was due to the FHFA prohibiting F&F from making most loan purchases deemed not to have “qualified mortgage” status.

[27] My view behind this characterization comes primarily from the writings and views he expressed while he worked at the Cato Institute, a self-described free-market research organization.

[28] His directorship and my time as Freddie Mac CEO overlapped for a few months in 2019. He did not elaborate on why CRT would be effective only until the conservatorship ended.

[29] April 2019 to June 2021.

[30] The CCF had been an unofficial capital system for use during conservatorship. The ERCF, by contrast, was an official regulatory rule, replacing a predecessor that the FHFA had suspended in 2008.

[31] This means it contained many features that were not risk-based. As but one example of specific relevance to CRT, it contained an arbitrary minimum (also called a floor) that simply overrode the reduction of risk by a CRT transaction.

[32] Market disruptions during the height of COVID-19 also contributed to this decline.

[33] Director Calabria’s successor, Sandra Thompson, revised the ERCF in some targeted ways to reduce the impact of the underestimation. This significantly contributed to the first phase of the hollowing out ending.

[34] This adjustment was based on the belief that, because prices could potentially decline further, F&F’s guarantees to MBS investors could generate additional credit losses.

[35] This reflects the risk transfer percentage moving down from the 60 percent range to the 20 percent range (The 13 percent range for 2025 is impacted by it being a partial year calculation.)

[36] Any material change to a regulatory capital system will likely produce a fair amount of political activity by various interest groups both for and against the change. This is one reason why such changes are not made frequently and only if the benefit of the changes is material in impact.

[37] It is worth noting that, if and when F&F are released from conservatorship, it is possible that the FSOC will officially designate the two companies as “systemically important financial institutions.” If that happens, the Federal Reserve will become an additional regulator, and it is impossible to estimate what that might mean for the three risks specified in this two-part series. For example, the Federal Reserve might establish an investment portfolio limit, it might mandate greater transfer of credit risk, and so on.